gh-117376: Fix off-by-ones in conversion functions

[Fidget-Spinner]

This wasn't caught because the specializations are currently not tested in the free-threaded builds.

To test this, apply the following patch to main:

diff --git a/Python/ceval_macros.h b/Python/ceval_macros.h
index 9e1540674d4..7ca2400529d 100644
--- a/Python/ceval_macros.h
+++ b/Python/ceval_macros.h
@@ -441,28 +441,13 @@ do { \
 /* How much scratch space to give stackref to PyObject* conversion. */
 #define MAX_STACKREF_SCRATCH 10
 
-#ifdef Py_GIL_DISABLED
 #define STACKREFS_TO_PYOBJECTS(ARGS, ARG_COUNT, NAME) \
     /* +1 because vectorcall might use -1 to write self */ \
     PyObject *NAME##_temp[MAX_STACKREF_SCRATCH+1]; \
     PyObject **NAME = _PyObjectArray_FromStackRefArray(ARGS, ARG_COUNT, NAME##_temp + 1);
-#else
-#define STACKREFS_TO_PYOBJECTS(ARGS, ARG_COUNT, NAME) \
-    PyObject **NAME = (PyObject **)ARGS; \
-    assert(NAME != NULL);
-#endif
 
-#ifdef Py_GIL_DISABLED
 #define STACKREFS_TO_PYOBJECTS_CLEANUP(NAME) \
     /* +1 because we +1 previously */ \
-    _PyObjectArray_Free(NAME - 1, NAME##_temp);
-#else
-#define STACKREFS_TO_PYOBJECTS_CLEANUP(NAME) \
-    (void)(NAME);
-#endif
+    // _PyObjectArray_Free(NAME - 1, NAME##_temp);
 
-#ifdef Py_GIL_DISABLED
 #define CONVERSION_FAILED(NAME) ((NAME) == NULL)
-#else
-#define CONVERSION_FAILED(NAME) (0)
-#endif

Note that Python fails to build on main, but passes on this branch.

• Issue: Implement deferred reference counting in free-threaded builds #117376

[rruuaanng]

I think if this patch fails to pass CI on the master branch, it may cause the code base to fail testing after merging.
You should probably try it through CI (Hmm, because I had a performance optimization patch that did that before, and it got shut down).

[Fidget-Spinner]

@rruuaanng the failures are unrelated. If you look at logs it's because it cant fetch LLVM for JIT from the package manager.